Abstract:The ensemble square-root filter (EnSRF) is a kind of deterministic ensemble-based data assimilation method,and is used in a growing number of research fields and applications.Ensemble methods,compared with variational methods,require little expert knowledge for the development of tangent linear and adjoint versions of models and forward observation operators,the background-error covariances are flow dependent,and they can be combined with the ensemble forecast.At the same time,EnSRF based on the traditional ensemble Kalman filter update equation,ameliorates the impacts of sampling errors introduced by adding random perturbations to the observations.Furthermore,the computational cost of the method is relatively lower compared with that of other deterministic ensemble methods;plus,EnSRF is easy to code and implement.Because of these advantages,EnSRF has become a hot topic in research and applications related to data assimilation.Owing to its high temporal and spatial resolution,Doppler weather radar has become the most effective method in monitoring and providing warnings for severe convective weather.The assimilation of Doppler radar data is therefore important for the improvement of storm-scale numerical weather prediction.To retrieve dynamically consistent wind,thermodynamic and microphysical fields from radar radial velocity and reflectivity,advanced data assimilation methods are required.According to operational needs,a WRF-EnSRF system for storm-scale assimilation was constructed in previous work.The study involved developing key assimilation techniques of the WRF-EnSRF system,and introduced adaptive localization and adaptive covariance inflation error correction algorithms to help filters to tolerate errors from many sources,including sampling errors,model errors and fundamental inconsistencies between the filter assumptions and reality,which lead to insufficient variance in ensemble state estimates.During the ideal storm tests,the results showed the characteristics and a good performance level of the adaptive algorithms developed,and a better assimilation scheme was obtained.During the real tests of assimilating Doppler radar data,the adaptive localization and adaptive covariance inflation introduced demonstrated it was possible to take into consideration many complex factors of influence.In the present study,based on the assumption that a multi-scheme ensemble forecast that combines different microphysical parameterization schemes may significantly improve the performance of EnKF,as opposed to using a single scheme,the WRF-EnSRF system is examined to assimilate the simulated radar data of a typical super storm that occurred on 20 May 1977 in Oklahoma city,USA.Based on the self-developed WRF-EnSRF data assimilation system,this study assimilates the simulated Doppler radar data and discusses the impact of microphysical schemes and the uncertainty of their parameters on the performance of EnSRF data assimilation,and uses the improved scheme in a series of comparison tests involving the assimilation of simulated radar data.Different mixes of microphysical schemes and perturbations of microphysical parameters are involved in the experiments.The overall goal of the research was to develop an EnSRF data assimilation system and to investigate its ability in radar data assimilation for storm-scale numerical weather prediction.The results show that,in the absence of model error,using a single microphysical scheme with its parameters perturbed,retreives the main features of the storm better than without the perturbed parameters.This difference,especially for the spatial distribution of most variables in the analysis,becomes more significant in the presence of model error.In this case,synchronouly involving a mix of microphysical schemes and the perturbation of their parameters produces convective clouds in the analysis that are better than without any one of these two approaches.With both approaches,data assimilation produces the best result among all the experiments with model errors;the main features of the storm are reasonably retrieved.Meanwhile,results also show that the range of parameter perturbation has to be small enough to produce an optimal analysis.